deliberation-debate-red-teaming
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No security vulnerabilities, malicious code, or suspicious patterns were found in the skill. The instructions and resources focus exclusively on providing a structured methodology for adversarial review and strategic deliberation.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process and critique user-provided proposals. While this involves ingesting untrusted data (Ingestion points: SKILL.md Step 1), the skill lacks tool-based capabilities (such as network or file system access) that could be abused via prompt injection (Capability inventory: None). Furthermore, the core purpose of the skill—adversarial review—is a safety-positive behavior that helps surface risks in processed content. (Boundary markers: Absent, Sanitization: Absent).
Audit Metadata